Anthropic's Claude Sonnet 5.5 Is Out, Beats Opus 5.5 at Coding for Half the Price

What happened

Anthropic released Claude Sonnet 5.5, a mid-tier model that outperforms its own flagship Opus 5.5 on Terminal-Bench 4.0 for coding and costs half as much per token, though an independent tester found it burns more tokens than any model measured.

Confirmed

Global impact / market context

Cheaper tokens could lower the cost of building software with artificial intelligence, which may let companies spend more on development. But heavier token use means the real bill could be higher than the headline price suggests, squeezing budgets.

Analyst inference

Anthropic competes with other artificial intelligence model makers on price and performance. A mid-tier model outperforming a flagship suggests rapid improvement, which may push rivals to cut prices or release better models, affecting revenue and spending across the industry.

Analyst inference

What to watch

1. Anthropic claims Sonnet 5.5 beats Opus 5.5 on Terminal-Bench 4.0 and costs half as much per token, so watch whether those two figures hold up in wider testing. Confirmed
2. Watch whether Anthropic or rivals adjust token pricing or release competing models, which would change costs for companies that buy artificial intelligence services. Proposed
3. Watch whether heavy token use raises real customer bills, which could slow adoption if buyers find the cheaper model costs more in practice. Analyst inference

Evidence

• decrypt.co